Memory corruption issues in 18.42.2 on MCU1 cars?

[verygreen]

So... I got a memory controller error in my logs after upgrading to 18.42.2, looks like this:

[  124.827125] tegra-mc 7000f000.memory-controller: idxsrd: read @0x733efb20: EMEM address decode error (EMEM decode error)

And then I started hearing from people that they get glitched fonts/display issues and I think these are connected. With 18.42.2 apparently being a pretty wide release, I am sure other must be seeing these too?

So... If you saw anything like that or something related, please reply here. I am going to add a sample picture once I am able to procure one.

There's a sister post in the S forum to increase visibility

 

[verygreen]

here are some samples.

 

[scottf200]

Perhaps fixed in .3 or .4 that has gone out pretty fast. -- .4 seemed to really go out to M3s tho. (TeslaFI.COM fleet sample 3000ish cars)

 

[verygreen]

I dunno, my car did not get the .3 yet

 

[Shaz]

Yup got something similar as well. Did a full reboot and it went away. Haven't had it occur since.

 

[chdavis]

Yup got something similar as well. Did a full reboot and it went away. Haven't had it occur since.

Same here

 

[Twoeagle]

Same,same. 42.2.

 

[TB205gti]

Yup.. 48.12.1, starting to get corrupted fonts in charging screen. :/

..And a *sugar* ton o the following in dmesg

tegra-mc 7000f000.memory-controller: idxsrd2: read @0x72665aa0: EMEM address decode error (EMEM decode error)

 

[verygreen]

Yup.. 48.12.1, starting to get corrupted fonts in charging screen. :/

..And a *sugar* ton o the following in dmesg

tegra-mc 7000f000.memory-controller: idxsrd2: read @0x72665aa0: EMEM address decode error (EMEM decode error)

I noticed that these also come out in force when running mame...

 

[Ruffles]

Mine was corrupted as well but a full reboot of everything fixed it.

 

[TB205gti]

I noticed that these also come out in force when running mame...

Hmm interesting.. I did try mame a few days ago. utterly useless feature

I'll keep a look out for the errors, but interesting that mame might be the trigger..

 

[verygreen]

Hmm interesting.. I did try mame a few days ago. utterly useless feature

I'll keep a look out for the errors, but interesting that mame might be the trigger..

it's not the trigger,but whatever it does (high cpu usage?) seems to trigger the messages too, they can be triggered without mame as well, but kind of randomly.

 

[TB205gti]

it's not the trigger,but whatever it does (high cpu usage?) seems to trigger the messages too, they can be triggered without mame as well, but kind of randomly.

Do we know what memory (idxsrd2?) it is - I hope it's not the eMMC that is dying..

 

[Casavant]

So... I got a memory controller error in my logs after upgrading to 18.42.2, looks like this:

[  124.827125] tegra-mc 7000f000.memory-controller: idxsrd: read @0x733efb20: EMEM address decode error (EMEM decode error)

And then I started hearing from people that they get glitched fonts/display issues and I think these are connected. With 18.42.2 apparently being a pretty wide release, I am sure other must be seeing these too?

So... If you saw anything like that or something related, please reply here. I am going to add a sample picture once I am able to procure one.

There's a sister post in the S forum to increase visibility

Had the same problem on charging related screens... Full reboot didn't help, full reboot after next update did take cate of it

 

[Peteski]

Yup.. 48.12.1, starting to get corrupted fonts in charging screen. :/

+1 on corrupted charging screen display. I get it maybe every 2nd or 3rd time I plug in and then it usually clears on reboot or power cycle. Media player (Spotify) has also been a nightmare on this firmware, freezing on a daily basis.

 

[docBliny]

Been getting same. Font and other UI glitches in various settings. Homelink stops working. A reboot fixes it for a day or two. 2018.48.12.1.

 

[dgatwood]

Do we know what memory (idxsrd2?) it is - I hope it's not the eMMC that is dying..

Although the comments in the driver source code are confusing as heck, as best I can tell, idxsrd2 is TEGRA_SWGROUP_NV2, which is a group that AFAICT represents memory or storage that is part of the Tegra SoC itself in one specific model of Tegra SoC. I'm guessing from the name that it is a GPU.

The MMC hardware, as best I can tell, would be any of TEGRA_SWGROUP_SDMMC1A through TEGRA_SWGROUP_SDMMC4A.

Given how many people are seeing it, this is likely a kernel/driver bug, rather than a hardware problem.